Fireflies MCP Server: How to Connect Your Meeting Data to Claude, ChatGPT, and More

Your meetings hold the context your AI tools need, but it stays buried in transcripts, summaries, and action items from weeks of calls. The Fireflies MCP server connects that meeting data directly to Claude, ChatGPT, Cursor, Devin, and other AI tools through one secure connection. You ask about any past call in plain language, without copying notes between apps.

Fireflies built the connector on the Model Context Protocol (MCP), the open standard that links AI applications to outside data. The server was first introduced in mid-2025, and on August 21, 2025 it became the first AI meeting tool added to Anthropic's Claude Connectors Directory. 

The launch is backed by strong early adoption. During the beta that began June 25, 2025, 55% of users returned for multiple sessions and weekly usage grew 20%.

Following this directory debut, it is now generally available. If you’d like to learn more about the Fireflies MCP server, this guide explains what MCP is, what it can do, and how to set up one across your tools.F

What is the Fireflies MCP server?

The Fireflies MCP server is an official connector that links your Fireflies meeting data to AI tools through the Model Context Protocol. It runs at https://api.fireflies.ai/mcp, authenticates with OAuth, and lets tools like Claude and ChatGPT read and act on your transcripts, summaries, and action items.

Fireflies runs this official server itself, so you connect straight to the source. This mechanism eliminates the need to route your meeting data through community servers or outside middleware platforms that broker MCP access. It is also unrelated to Firefly (firefly.ai), a separate cloud and DevOps tool with a similar name.

Anthropic lists it in the Claude Connectors Directory, and the same remote URL works across Claude, ChatGPT, Cursor, Devin, and other MCP clients. You copy the URL once from your Fireflies settings, then authorize it in your AI tool. Full setup lives in the MCP configuration docs, and the rest of this guide walks each client step by step.

What is MCP (Model Context Protocol)?

Model Context Protocol is an open standard that lets AI applications connect to outside data and tools through one common interface. Before MCP, every AI tool needed a custom integration for every data source. MCP replaced that with a single protocol, so any compliant tool can reach any compliant server.

The standard moved fast. Anthropic introduced MCP in November 2024. OpenAI adopted it the following March, and Google confirmed Gemini support a month later. On December 9, 2025, Anthropic donated MCP to the Linux Foundation's new Agentic AI Foundation, alongside Block's goose and OpenAI's AGENTS.md. By then it had passed 10,000 active servers and 97 million monthly SDK downloads. The current specification, released July 28, 2026, reworked the protocol around a stateless core and added an extensions framework for capabilities like asynchronous tasks.

What can you do with Fireflies MCP?

Once connected, your AI tool treats your meeting history as a searchable knowledge base. What it does with that depends on your team:

• Sales teams surface objections, deal risks, and competitor mentions across recent calls. Ask "What were the common objections in this week's sales calls?"
• Product teams turn scattered user interviews into one clear view of what people asked for. Ask "Summarize all product feedback from user interviews this month."
• Customer success teams track how sentiment and recurring issues shift over time. Ask "How has customer sentiment changed over the past quarter?"
• Engineering teams keep build decisions tied to the conversation that shaped them. Ask "What were the security concerns raised in last week's architecture review?"

Much of this mirrors what AskFred already does inside Fireflies. The difference is where you work. The MCP server brings the same recall into Claude, ChatGPT, or your coding tool, so you never leave the app you are in.

Fireflies MCP tools

The server exposes two kinds of tools. Read tools pull meeting data. Action tools change it. Here is the full set.

Two of these, fireflies_search and fireflies_fetch, are experimental and rolling out gradually, so they may not appear for every account yet. The complete reference, with parameters and return fields, lives in the full MCP tools reference.

Which tools you see depends on how you connect. The OAuth connector used by most AI tools currently surfaces a subset of these, focused on reads like user, transcript, and search. The action tools are available through the server itself. If a tool you expect is missing, check your connection method first.

How to set up the Fireflies MCP server

Setup follows the same shape in almost every tool. You copy one URL from Fireflies, add it as an MCP connector in your AI tool, and sign in with OAuth. The whole thing takes about a minute per client.

Start in Fireflies. Log in, open the Integrations page, select the MCP category, and choose the tool you want to connect. That takes you to Settings, where you copy the MCP Server URL (it reads https://api.fireflies.ai/mcp). Keep it handy for the client steps below. Tools that already list Fireflies in their connector directory, like Claude, skip the copy step. For a walkthrough with screenshots, see the step-by-step connector guide.

Connect Fireflies to Claude in under 2 minutes. OAuth, no config files, works across every Claude surface. Connect Fireflies to Claude.

Connect Fireflies to Claude

Claude lists Fireflies in its Connectors Directory, so you add it in four steps.

1. In Claude, open Settings, then Connectors.
2. Click Browse connectors and find Fireflies.
3. Click Connect, then sign in to Fireflies with Google or Microsoft.
4. Approve access. Claude can now read your meetings.

The connection carries across Claude Web, Desktop, Mobile, and Claude Code, so you authorize once and it follows you. For background on what the connector pulls in, see the Fireflies connector for Claude. You'll need a paid Claude plan (Pro, Team, or Enterprise) with access to the Connectors feature.

Connect Fireflies to ChatGPT

Connecting Fireflies to ChatGPT uses OAuth, with no API key required. You need a ChatGPT Plus, Team, or Enterprise account with access to the Connectors feature.

1. Go to ChatGPT Settings, then Connectors, and click Create.
2. Enter the connector details. Name it Fireflies, set the MCP Server URL to https://api.fireflies.ai/mcp, and choose OAuth for authentication.
3. Check "I trust this application."
4. Click Create.
5. Complete the OAuth sign-in when prompted.

Once connected, you can query your Fireflies meeting data inside ChatGPT.

Connect Fireflies to Devin

Fireflies is the first AI meeting tool officially listed on Devin's MCP marketplace, so setup runs through Devin's built-in marketplace.

1. In Devin, go to Settings, then MCP Marketplace, and find Fireflies in the featured integrations.
2. Click the Fireflies tile, then click Enable and check the box to confirm.
3. Follow the prompts to authorize Devin's access to your Fireflies data with OAuth.
4. Test the connection by asking Devin something like "Show me my recent meetings."

Once connected, Devin can reference the meetings that shaped your requirements while it writes code. For the full walkthrough, see the Devin setup guide.

Connect Fireflies to Cursor and other MCP clients

Any tool that supports remote MCP servers connects with the same URL and a bearer token.

1. In Cursor, open Settings, then MCP, and add a new server.
2. Set the server URL to https://api.fireflies.ai/mcp.
3. Add an Authorization header with the value Bearer YOUR_API_KEY, using your Fireflies API key.
4. Save, then start querying your meetings.

Other clients follow the same pattern, a server URL plus a bearer token or OAuth. Some use a different label for the URL field. Windsurf, for example, calls it serverUrl, but the value is identical.

Other tools you can connect

Fireflies keeps adding MCP connectors, so the same meeting data now reaches design, build, and agent tools beyond the chat and coding clients above. Each uses the same setup flow. Pick it from the MCP category in Fireflies, then authorize it with OAuth.

• Figma lets you turn meeting discussions into wireframes and prototypes and reference decisions while you design.
• Lovable turns planning calls and product discussions into working applications (requires Lovable's Business plan for custom MCP connectors).
• Bolt builds full-stack apps from the requirements and customer feedback raised in meetings.
• Magic Patterns generates interactive UI prototypes and user flows from your design reviews.
• Poke answers questions about past meetings from your messaging apps.
• Replit turns meeting takeaways into apps you build and deploy in one place.
• Manus AI turns action items and follow-ups into tasks your AI agent can complete.

New connectors land regularly, so check the connector guide for the current list.

Claude Desktop with your API key

Most people connect Claude through OAuth, but Claude Desktop also supports a manual setup with your Fireflies API key. Use this if you prefer a local config or want to authenticate with a key instead.

First, get your key. In Fireflies, open Settings, then Developer Settings, and copy your API key. Then paste this block into your claude_desktop_config.json file.

{
  "mcpServers": {
    "fireflies": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.fireflies.ai/mcp",
        "--header",
        "Authorization: Bearer YOUR_API_KEY_HERE"
      ]
    }
  }
}

Replace YOUR_API_KEY_HERE with your actual key, save the file, and restart Claude Desktop. The connector loads on the next launch.

Is the Fireflies MCP server free?

Yes. The Fireflies MCP server is available on every plan, including the free tier, and connecting it costs nothing beyond your Fireflies subscription. Free accounts get API access and 20 AI credits, which power AskFred and other AI features. Business and Enterprise plans add more credits and higher rate limits.

Usage runs against the same rate limits as the Fireflies API, and those limits apply to the MCP server too. Free and Pro accounts get 50 requests per day. Business and Enterprise move to 60 requests per minute.  Business and Enterprise move to 60 requests per minute. For everyday meeting questions that ceiling is generous, and the higher tiers are where heavy automated use pays off.

Security and privacy

Connecting your meetings to an outside AI tool raises a fair question about where the data goes. Two things stay constant. Fireflies has never trained its AI models on customer meeting data, and the connector respects your existing Fireflies permissions, so a connected tool reaches only the meetings your account can already see. Access runs through OAuth, and data moves over Fireflies' standard secure transfer.

Retention is where the boundary matters. Fireflies maintains a strict Zero-Day Retention policy with the sub-processors behind its core features, including summaries, AskFred, and transcription. That policy covers processing inside Fireflies. It does not extend to an outside AI tool you connect through MCP. Once your meeting data reaches Claude, ChatGPT, or another client, that platform handles it under its own terms, and Fireflies cannot guarantee how long it is stored or whether it is used to train that platform's models. If your organization has strict data rules, connect MCP tools only where external AI processing is allowed, and rely on Fireflies' built-in features or API integrations when you need Zero-Day Retention to hold.

On compliance, SOC 2 Type II and GDPR apply across every tier, including Free. HIPAA compliance is available on Enterprise under a BAA.

Frequently asked questions

What is the Fireflies MCP server?

The Fireflies MCP server is the official connector that makes your Fireflies meeting data available to AI tools through the Model Context Protocol. It runs at https://api.fireflies.ai/mcp and authenticates with OAuth.

How do I connect Fireflies to Claude?

To connect Fireflies to Claude, open Claude's Settings, go to Connectors, browse for Fireflies, click Connect, and sign in with Google or Microsoft. The connection then carries across Claude Web, Desktop, Mobile, and Claude Code.

How do I set up the Fireflies MCP server?

You set up the Fireflies MCP server by copying its URL from Fireflies Settings, then MCP & Dev Tools, and adding it as an MCP connector in your AI tool. Authorize with OAuth and you are connected, with no API key required for most tools.

What can you do with Fireflies MCP?

With Fireflies MCP you can ask an AI tool questions about past meetings, pull summaries and action items, search across calls by topic or speaker, and turn discussions into documents, code, or tasks. It all runs in plain language.

Is the Fireflies MCP server free?

The Fireflies MCP server is free on every plan, including the free tier, at no cost beyond your Fireflies subscription. Free and Pro accounts get 50 API requests per day, and Business and Enterprise move to 60 per minute.

How do I use Fireflies with ChatGPT?

To use Fireflies with ChatGPT, open Settings, then Connectors, and click Create. Name the connector Fireflies, set the MCP Server URL to https://api.fireflies.ai/mcp, choose OAuth, and complete the sign-in. It works on ChatGPT Plus, Team, and Enterprise plans with the Connectors feature.

Does Fireflies work with Cursor?

Fireflies works with Cursor and any other client that supports remote MCP servers. Add a new server in Cursor's MCP settings, point it at https://api.fireflies.ai/mcp, and include an Authorization: Bearer header with your API key.

Does Fireflies work with Devin?

Fireflies works with Devin, and it was the first AI meeting tool listed on Devin's MCP marketplace. Add it from the marketplace, authorize with OAuth, and Devin can reference your meetings while it writes code.

Is the Fireflies MCP server secure?

The Fireflies MCP server is secure by design. It uses OAuth, respects your existing Fireflies permissions, and never lets a connected tool see meetings your account cannot. Fireflies also does not train its models on customer data.

Does Zero-Day Retention apply to MCP connectors?

Zero-Day Retention does not apply to MCP connectors once your data reaches an outside AI tool. ZDR covers processing inside Fireflies, so after data moves to Claude, ChatGPT, or another client, that platform handles it under its own retention and privacy terms.

Conclusion

Your meeting history is one of the richest sources of context your AI tools could use, and until now it mostly sat locked inside Fireflies. The MCP server changes that. One connection puts your transcripts, summaries, and decisions inside Claude, ChatGPT, Cursor, Devin, and any other tool that supports MCP. This setup takes about a minute via OAuth, and Fireflies never trains its AI models on your conversations. It sits alongside the 100+ integrations Fireflies already connects to across your stack. Start on the free plan, connect your first tool, and ask it something you would normally dig through your notes to answer.